Towards End-to-End Synthetic Speech Detection (2021-06-11T00:00:00.000000Z)

TL;DR

It is shown that by only using standard components, a light-weight neural network could outperform the state-of-the-art methods for the ASVspoof2019 challenge, and the proposed model is termed Time-domain Synthetic Speech Detection Net (TSSDNet), having ResNet- or Inception-style structures.

Abstract

The constant Q transform (CQT) has been shown to be one of the most effective speech signal pre-transforms to facilitate synthetic speech detection, followed by either hand-crafted (subband) constant Q cepstral coefficient (CQCC) feature extraction and a back-end binary classifier, or a deep neural network (DNN) directly for further feature extraction and classification. Despite the rich literature on such a pipeline, we show in this paper that the pre-transform and hand-crafted features could simply be replaced by end-to-end DNNs. Specifically, we experimentally verify that by only using standard components, a light-weight neural network could outperform the state-of-the-art methods for the ASVspoof2019 challenge. The proposed model is termed Time-domain Synthetic Speech Detection Net (TSSDNet), having ResNet- or Inception-style structures. We further demonstrate that the proposed models also have attractive generalization capability. Trained on ASVspoof2019, they could achieve promising detection performance when tested on disjoint ASVspoof2015, significantly better than the existing cross-dataset results. This paper reveals the great potential of end-to-end DNNs for synthetic speech detection, without hand-crafted features.

Authors

Haijian Zhang

2 papers

Guang Hua

1 papers

A. Teoh

1 papers

TL;DR

Abstract

Authors

References48 items

End-to-End anti-spoofing with RawNet2

Replay and Synthetic Speech Detection with Res2Net Architecture

FastSpeech 2: Fast and High-Quality End-to-End Text to Speech

Assessing the Scope of Generalized Countermeasures for Anti-Spoofing

ASVspoof 2019: A large-scale public database of synthesized, converted and replayed speech

Replay Spoofing Countermeasure Using Autoencoder and Siamese Network on ASVspoof 2019 Challenge

Extraction of Octave Spectra Information for Spoofing Attack Detection

Understanding and Visualizing Raw Waveform-Based CNNs

Long Range Acoustic Features for Spoofed Speech Detection

Detecting Spoofing Attacks Using VGG and SincNet: BUT-Omilia Submission to ASVspoof 2019 Challenge

STC Antispoofing Systems for the ASVspoof2019 Challenge

Ensemble Models for Spoofing Detection in Automatic Speaker Verification

ASVspoof 2019: Future Horizons in Spoofed and Fake Audio Detection

ASSERT: Anti-Spoofing with Squeeze-Excitation and Residual neTworks

Spoofing Detection in Automatic Speaker Verification Systems Using DNN Classifiers and Dynamic Acoustic Features

Conv-TasNet: Surpassing Ideal Time–Frequency Magnitude Masking for Speech Separation

Synthetic speech detection using fundamental frequency variation and spectral features

Voice Impersonation Using Generative Adversarial Networks

Neural Voice Cloning with a Few Samples

mixup: Beyond Empirical Risk Minimization

End-to-End convolutional neural network-based voice presentation attack detection

An Exemplar-Based Approach to Frequency Warping for Voice Conversion

Constant Q cepstral coefficients: A spoofing countermeasure for automatic speaker verification

Long-Term Spectral Statistics for Voice Presentation Attack Detection

ResNet and Model Fusion for Automatic Spoofing Detection

Deep Feature Engineering for Noise Robust Spoofing Detection

Cochlear Filter and Instantaneous Frequency Based Features for Spoofed Speech Detection

Spoofing Speech Detection Using Modified Relative Phase Information

Spectral Features for Synthetic Speech Detection

Significance of Source–Filter Interaction for Classification of Natural vs. Spoofed Speech

Generalization of spoofing countermeasures: A case study with ASVspoof 2015 and BTAS 2016 corpora

An Investigation of Deep-Learning Frameworks for Speaker Verification Antispoofing

Spoofing speech detection using temporal convolutional neural network

Deep features for automatic spoofing detection

Synthetic speech detection using phase information

Deep Residual Learning for Image Recognition

Multi-Scale Context Aggregation by Dilated Convolutions

ASVspoof 2015: the first automatic speaker verification spoofing and countermeasures challenge

Toward a Universal Synthetic Speech Spoofing Detection Using Phase Information

Adam: A Method for Stochastic Optimization

Going deeper with convolutions

Very Deep Convolutional Networks for Large-Scale Image Recognition

Speech Synthesis Based on Hidden Markov Models

Significance of Subband Features for Synthetic Speech Detection

A Cross-Database Study of Voice Presentation Attack Detection

A comparison of features for synthetic speech detection

ASVspoof 2015: the first automatic speaker verication spoofing and countermeasures challenge

, challenge

Field of Study

Journal Information

Name

Page

Volume

Venue Information

Name

Type

URL

Alternate Names